Objective: "The elderly" is an often used but poorly defined descriptor of surgical patients. Investigators have used varying subjectively determined age cutoffs to report outcomes in the elderly. We set out to use objective outcomes data to determine the "at-risk" elderly population.
Patients: 129,331 patients identified from the ACS-NSQIP database (2005-2010) undergoing major gastrointestinal resections.
Outcome: Mortality.
Statistical methods: Locally weighted regression was used to fit the trend line of mortality over age. Receiver operating characteristic analysis was used to identify the "predictive age" for mortality.
Results: Mortality steadily increases with age. On receiver operating characteristic analysis, there is a nonlinear transition zone (50-75 years of age) flanked by 2 linear zones on either end. The younger linear zone showed a low mortality increase (0.5% per decade). Larger mortality increase with age (5.3% per decade) was observed at the older age end. Similar patterns were observed for large-volume surgical subtypes, with clustering of a "critical age" beyond which mortality increases dramatically at 75 ± 2 years. Receiver operating characteristic analysis identified the "optimum age" for mortality being 68.5 years (area under the curve = 0.72, sensitivity = 66.6%, and specificity = 65.5%).
Conclusions: Mortality risk for major gastrointestinal surgical resections starts increasing at 50 years of age, and at 75 years of age, it starts increasing very rapidly. The optimum age of 68.5 years predicts mortality with the best combination of sensitivity and specificity. These ages should be used to standardize outcome data and focus perioperative resources to improve outcomes.